Вы находитесь на странице: 1из 6

# Hindawi Publishing Corporation

## Journal of Inequalities and Applications

Volume 2011, Article ID 721827, 6 pages
doi:10.1155/2011/721827

Research Article
A Sharp Double Inequality for Sums of Powers

Vito Lampret
Department of Mathematics and Physics (KMF), Faculty of Civil and Geodetic Engineering (FGG),
University of Ljubljana (UL), 1000 Ljubljana, Slovenia

## Received 26 September 2010; Revised 7 December 2010; Accepted 11 January 2011

any medium, provided the original work is properly cited.

It is established that the sequences n → Sn : nk1 k/nn and n → ne/e − 1 − Sn are strictly
increasing and converge to e/e − 1 and ee  1/2e − 13 , respectively. It is shown that there holds
the sharp double inequality 1/e−1·1/n e/e−1−Sn < ee1/2e−13 ·1/n, n ∈ .

1. Introduction
The proof of the equality

n  n
 k e
lim  , 1.1a
n→∞
k1
n e−1

n−1  n
 k 1
lim  , 1.1b
n→∞
k1
n e−1

## was based on the equations n1−k · nn − 1 · · · n − k  2  1 − 1/n1 − 2/n · · · 1 − k − 2/n 

1  O1/n with the false hypothesis that big O is independent of k see 1, pages 63-64 and
2, pages 54-55. Deriving 1.1b the author used the Euler-Maclaurin summation formula
and a generating function for the Bernoulli numbers.
2 Journal of Inequalities and Applications

e/(e − 1)
1.5

0.5

10 20 30 40 50
n
Figure 1: The graph of the sequence n → Sn ≡ k1 k/nn .

Subsequently, Spivey published the correction of his demonstration as the Letter to the
Editor 2. Additionally, Holland 3 published two diﬀerent derivations of 1.1a in the same
issue as Spivey’s correction appeared.
In this note, using only elementary techniques, we demonstrate that the sequence Sn
is strictly increasing and that 1.1a holds; in addition, we establish a sharp estimate of the
rate of convergence.

2. Monotone Convergence

The formula 1.1a is illustrated in Figure 1, where the sequence n → Sn : nk1 k/nn is
depicted. Its monotonicity is seen very clearly.
To prove that the sequence Sn n∈ is strictly increasing, we change the order of
summation

n  n
 n 
  n  
k n−j n −j n
Sn ≡ ≡ ≡1 1 . 2.1
k1
n j0
n j1
n

## Now, consider the function t → Ex, t : 1  x/tt which is, for x /

 0, strictly increasing on
the open interval − min{0, x}, ∞ and limt → ∞ Ex, t  supt>|x| Ex, t  ex , for any x ∈ Ê
4, page 42. Consequently, the sequence Snn∈ is strictly increasing. We use Tannery’s
theorem for series see 5 or 6, item 49, page 136 to determine its limit.
Lemma 2.1 Tannery. Let a double sequence j, n → zj n of complex numbers satisfy the
following conditions:

1 The finite limit z∞ j : limn → ∞ zn j exists for every fixed j ∈ Æ .
2 There exists a sequence of positive constants M1 , M2 , M3 , . . . such that |zn j| ≤ Mj for

every j, n ∈ Æ × Æ satisfying the estimate j ≤ n, and the series ∞ j1 Mj converges.
(In [6, item 49, page 136], we have the stronger supposition that |zn j| ≤ Mj for all
j, n ∈ Æ × Æ .)

Then we have


n   ∞  
lim zn j  z∞ j . 2.2
n→∞
j1 j1
Journal of Inequalities and Applications 3

Proof. Let all the conditions of the Lemma be satisfied and ε ∈ Ê be given. Then we estimate

|z∞ j| ≤ Mj for j ∈ Æ and ∞ jmε 1 Mj < ε/3 for some mε ∈ Æ . Moreover, for any
j ∈ {1, . . . , mε }, also |z∞ j − zn j| < ε/3mε  for n ≥ nε j at some nε j ∈ Æ . Thus, for
n ≥ nε : max1≤j≤mε nε j, we estimate

 
  mε
∞   n         
∞    
n   
 z∞ j − zn j  ≤ z∞ j − zn j   z∞ j   zn j 
 
 j1 j1  j1 jmε 1 jmε 1
2.3
ε 
∞ 
n
ε ε ε
< mε ·  Mj  Mj <    ε.
3mε jmε 1 jmε 1
3 3 3

Now, using 2.1 and putting zn j  1  −j/nn and z∞ j  e−j into Tannery’s
Lemma, we obtain



e
lim Sn  1  e−j  . 2.4
n→∞
j1
e−1

## 3. The Rate of Convergence

Referring to Figure 1, the convergence of the sequence Snn∈ appears to be rather slow.
The diﬀerence

e
Δn : − Sn 3.1
e−1

determines the sequence n → nΔn. Its graph, shown in Figure 2, suggests it is monotonic
increasing, which we will prove first.
Indeed, according to 3.1 and 2.1, we have

 n  

−j j n
Δn  e − 1−
j0 j0
n


n   ∞
 fn j  e−j 3.2
j1 jn1

n   e−n
 fn j  ,
j1
e−1

where

 n
x
fn x : e −x
− 1− x ∈ Ê 3.3
n
4 Journal of Inequalities and Applications

1
0.98
0.96
0.94
0.92

## Figure 2: The graph of the sequence n → nΔn.

and, for x 
/ 0, the sequence n → fn x is strictly decreasing and converges to zero 4, 4.
Thus, we have


n   e−n 
n−1  
nΔn  gn j  n  gn j  Cne−n 3.4
j1
e − 1 j1

with

e
gn x : nfn x, C . 3.5
e−1

To examine the monotonicity of the sequence n → nΔn, we study, using 3.3, 3.4
and 3.5, the diﬀerence n  1Δn  1 − nΔn, which is equal to
⎛ ⎞

n−1   
n−1  
⎝ gn1 j  gn1 n⎠  C · n  1e−n−1 − gn j − Cne−n
j1 j1


n−1      1 n  1 −n en −n
 gn1 j − gn j  n  1e−n −  e − e
j1 n  1n e − 1 e−1
3.6

n−1   
    e −n
 n  1fn1 j − nfn j  e − n  1−n
j1
e−1


n−1     
> nfn1 j − nfn j  0 > 0.
j1

Hence:

## The sequence n −→ nΔn is strictly increasing. 3.7

Next, we examine also the question of convergence of the above sequence. First,
referring to 3.3, 3.5, and 4, page 29, equation 16, there exists the limit

    e−j j 2  
g∞ j : lim gn j  j∈Æ . 3.8
n→∞ 2
Journal of Inequalities and Applications 5

## Moreover, according to 3.3, 3.5, and 4, 15, the estimates

  e−j j 2 n e−j j 2  
gn j < · ≤ · 1j 3.9
2 n−j 2

hold true for j ≤ n − 1. Additionally, gn n  ne−n , due to 3.3 and 3.5. Thus, the estimate

 
  j  1 j 2 −j
gn j ≤ Mj : ·e 3.10
2

## is being valid for n ∈ Æ and j ≤ n with

 
∞ ∞ j  1 j2
Mj  · e−j < ∞. 3.11
j1 j1
2

According to 3.8 and diﬀerentiating the appropriate power series resulting from the
geometric series, we obtain

∞    ∞ −j 2
e j e2  e
g∞ j   . 3.12
j1 j1
2 2e − 13

## Now, referring to 3.4 and 3.8–3.12, and applying Tannery’s Lemma—equation

2.2, with zn j ≡ gn j, we obtain the result

∞
  ee  1
lim nΔn  g∞ j  0  . 3.13
n→∞
j1 2e − 13

Therefore, using 3.1 and 3.7, we find the following sharp inequality

e ee  1 1
− Sn < · , 3.14
e−1 2e − 13 n

## true for every n ∈ Æ . In addition, we have also the estimate

 
e e 1
− Sn ≥ m − Sm · , 3.15
e−1 e−1 n

## valid for every m, n ∈ Æ such that n ≥ m.

6 Journal of Inequalities and Applications

We have ee  1/2e − 13  0.996147 . . ., and for the function P m : mΔm 
me/e − 1 − Sm we calculate P 1  0.581976 . . ., and P 999  0.995149 . . .. This way we
obtain simple and rather accurate estimates

1 e 1
0.581 · < − Sn < 0.996 · , for n ≥ 1,
n e−1 n
3.16
1 e 1
0.995 · < − Sn < 0.997 · , for n ≥ 1000.
n e−1 n

## Consequently, we get, for example, a simple double inequality

e 1 e 1
− < Sn < − , for n ≥ 1. 3.17
e−1 n e − 1 2n

Open Question. Are the sequences n → Sn and n → nΔn strictly concave?

Acknowledgments
The author wishes to express his sincere thanks to Prof. I. Vidav for some useful suggestions
and also to the referee whose comments made possible a significant improvement of the
paper.

References
1 M. Z. Spivey, “The Euler-Maclaurin formula and sums of powers,” Mathematics Magazine, vol. 79, pp.
61–65, 2006.
2 M. Z. Spivey, “Letter to the editor,” Mathematics Magazine, vol. 83, pp. 54–55, 2010.
 m
3 F. Holland, “limm → ∞ m k0 k/m  e/e − 1,” Mathematics Magazine, vol. 83, pp. 51–54, 2010.
4 V. Lampret, “Estimating powers with base close to unity and large exponents,” Divulgaciones
Matematicas, vol. 13, no. 1, pp. 21–34, 2005.
5 J. Tannery, Introduction à la Théorie des Fonctions d’une Variable, Sect. 183, Hermann, Paris, France, 1886.
6 T. Bromwich, Introduction to the Theory of Infinite Series, Chelsea, NY, USA, 3rd edition, 1991.